TensorPRO
TensorPRO builds trustworthy memory infrastructure for AI systems.
AI that loses context, hallucinates, or can't show where its knowledge came from can't be trusted with real work. As AI moves into regulated and high-stakes settings, organisations need to govern what their AI knows, trace it, audit it, and keep control of it.
Our product, the EDN Memory Engine is a governed long-term memory layer for assistants, agents or applications. Every memory carries a record of where it came from, and that provenance governs what your model may treat as fact. Retrieval is a deterministic pipeline, with no LLM on the retrieval path. EDN works with any LLM, and your memory is fully portable: switch models without losing a thing, or export it and take it with you. Your AI's memory stays yours.
In our sealed June 2026 evaluation (CONFIRM-1, LongMemEval, 500 questions), Claude Haiku 4.5 with EDN scored 85.4% (n=3 mean) against a 45.5% no-memory baseline.
We build for organisations where AI output is acted on: financial services, health, government, insurance, research, and the engineering teams building for them.
TensorPRO is a research-led team based in Brisbane, Australia. EDN is in early access through a REST API and web console.